Latest national poll median date: October 20
Projections reflect recent polling graciously made publicly available by pollsters and media organizations. I am not a pollster, and derive no income from this blog.

Wednesday, November 3, 2010

US Election Post Mortem

Relative to expectations, the midterm elections in the US were a mixed bag for both parties. Here is a tentative summary of what happened in Senate, House and Governor races.

Senate (51-53 Democrats, 47-49 Republicans)
- Harry Reid's relatively easy win in NV proved all pollsters in the last 3 weeks of the campaign wrong. It's not a big surprise since everyone knew it was going to be close, and President Obama significantly outperformed polls in NV in 2008. Still, the margin of victory (5%) is higher than what most would have predicted.
- CO and WA are also likely to remain Democratic, in which case the Dems would have 53 seats, making their majority defection-proof. The gap in the vote count is about 7,500 votes in CO and 14,000 votes in WA. CO looks to be the safer of the two: most votes have been counted, and almost all of the outstanding ones are in Democratic counties. Again, pollsters have been proven wrong here: no one gave the Democrat a lead in the last month of the campaign. There is a lot of counting left in WA, but King county, which includes Seattle, is a bit behind the rest of the state, which is a good sign for the Democrat.
- PA flipped to the Republicans, as expected. However, the margin there was only 2%, much less than the poll consensus. The margin in WI, which the GOP also took over as predicted, is also a bit lower than expected.
- Meanwhile, WV easily remained Democratic by 10%, a much larger difference than polls suggested.
- Overall, pollsters overestimated Republican support in almost all close Senate races, though NV was the only state where they were far off. This pales in comparison to their performance in 2006, when the consensus polling average was essentially on the dot in every close state.
- In 2012, 23 Democrats and 10 Republicans are scheduled to face an election. Even in a neutral environment, Democratic losses would be expected. The unresolved races in CO and WA are crucial: with 53 seats, the Dems have a fighting chance of maintaining their majority beyond 2012, but with just 51, that would be hard to do.

House (241-246 Republicans, 189-194 Democrats)
- In the House, 11 races remain uncalled, though only 5 are genuinely unsettled: VA-11 (Dems lead by 500 votes), KY-6 (D +600), IL-8 (R +800), WA-2 (R +1400), CA-11 (R +23). In each of these districts, only a few precincts are missing, except in WA-2, where due to postal voting, about 36% of the vote may be outstanding. CA-11 is almost certainly headed for a recount.
- Splitting the above 5 races evenly gives the GOP 243-244 seats, a gain of 65-66 relative to the 2008 election. This is the highest Republican total since the 1946 election, the biggest Republican seat increase since 1938, and the biggest Republican seat increase constituting a takeover since 1894! However, the last two Democratic Presidents facing a GOP takeover of a comparable magnitude, Truman and Clinton, were both comfortably reelected two years later.
- The Republican wave, while huge, was also not much bigger than expected: only 10 or so more seats than what most pundits called for. According to Nate Silver of FiveThirtyEight, the national popular vote was fairly well estimated by pollsters; the GOP swing was just a bit more efficient than expected.
- Relative to the 2008 election, the Democrats lost seats in 34 33 of the 46 states where they had at least one. The wave hit everywhere, but less hard in New England (where NH was the only state out of 6 to elect Republican representatives) and on the West Coast (where the Democrats lost only 2-4 1 districts). Outside the Northeast and the West Coast, the Democrats won 136 seats out of 274 in 2008, or about 50%. This time, they're down to about 86, or about 31%.
- There were only two Republican losses relative to 2008, both in very blue areas: Delaware, where the popular moderate GOP incumbent decided to run for Senate (and lost to a Tea Party candidate in the primary), and LA-2 (New Orleans), where in 2008, voters barely threw out a corrupt Democrat caught with $90,000 in his freezer. Also, the Dems took back HI-1, which had gone to the GOP in a special election where the Democratic vote was split. (Addendum: They also held NY-23, won in a special election last year.)

There are no surprises among the races that have already been called. The GOP picked up 10 seats mostly in very red states and the Midwest; this is going to help them for redistricting. They lost RI to an independent and CA (addendum: and HI) to a Democrat. Seven races remain outstanding and close:
- Likely GOP wins: ME (as expected), FL (was a tossup)
- Likely Democratic win: VT (as expected)
- CT: the GOP has a healthy 1% lead, but 10% of the vote, reportedly mostly from Democratic strongholds, remain outstanding. This race turned against the Democrat in the last 10 days, which polls did pick up.
- IL: the Democrat has a 0.3% lead, with a few precincts left. This is a surprise, as the GOP candidate was favored.
- MN: the Democrat has a 0.5% lead, with a few precincts left. This is closer than expected, though the Democrat may well hold on.
- OR: similar to CT: the GOP has a 1.1% lead, but 11% of the vote, mostly from around Portland, have not yet been counted. This one was a tossup.

Democrats outperformed expectations in the Senate, but underperformed in the House, while gubernatorial races were mixed. Why did the polls underestimate the Democrats in the Senate, but not the House? Here are a few possible explanations:
1. Senate races are more prominent. If you're a young or unenthusiastic voter filtered out by pollsters' likely voter screens, you're more likely to defy expectations and actually vote in a close Senate race than if the most important close race where you live is for the House.
2. It just so happens that the close Senate races were in states where pollsters had a GOP bias. Maybe the same pollsters would overestimate Democratic vote if the close races were elsewhere.
3. Pollsters are different for the House and the Senate, and it just so happens that Senate polling firms made errors in the same direction.
4. Undecided voters broke for the Democrats only in Senate races. Perhaps these voters really want to send a message to the Dems, but have no confidence in the GOP. After agonizing for a long time, they decide to split their ticket: send a message with their House vote, which they can change in 2 years, but not taking a risk with a somewhat unknown GOP Senate challenger, whom they can only throw out in 6 years.

I will post an update when most of the remaining races are resolved.

No comments: